I did an experiment a year or so ago where I compared the error in a Gamin handheld GPS unit to my Samsung Galaxy Nexus smart phone. I discovered that the Garmin is roughly twice as accurate (or more specifically has half the error) of the smart phone.
http://blog.oplopanax.ca/2012/11/mea...-gps-accuracy/

I've always been interested in comparing the performance of these devices, and under different circumstances. So I developed a web site to allow members of the community to upload their own GPS tracks, and analyse and compare the error in the device.


What you do:
* record a track while your GPS is stationary
* upload the track and some details of the experiment to the web site
* I will plot the distribution of the error and calculate the 50% and 95% confidence intervals.


http://gpserror.azurewebsites.net/

I would be interested in hearing from people on how this could be improved. If there is no interest. I'll eventually take it down.

Current plans for improvement
* comparison of error plots at same scale
* live gathering of data from a device

Help requested;
* I need a good data gathering system for iPad and iPhone that I can recommend to users.
* for Samsung and Andoid I used the MyTracks app, if anyone has a better suggestion I would love to hear it.

This is quite interesting, sometimes I forget to turn the GPS off after reaching a destination and end up with a wild psychedelic star shaped track around the stationary GPS. I have always assumed this is due to the "variability" noise that is added to the signal by the US military and not attributed to GPS receiver errors. Can you shed some light on this?

I've used Maprika on my iphone to record a track. It seemed ok. I didn't compare it to my garmin though.

http://www.maprika.com/

Its also available for android. Cool thing about this app is you can load a jpg map, geotag it and use your map for navigation! Its actually pretty accurate...

Remind your participants to include details of any tweaking they may have done to their smart phone to improve any error factors eg: firmware upgrades that resolve PDOP etc.

Also if this is for SAR use, bear in mind that individuals using their cell phones may be incorporating Google find application (which is common) into their location abilities and for older phones Cel-locate through tower triangulation. To get a perfect base line of your test, users should be providing location based upon GPS platform alone with 3rd party mapping software.

It's not for SAR use.

You're entirely right that phones in particular (both Android and iPhone) use various techniques for rough location including MAC addresses, ip geolocation and cell towers. None of these techniques is as accurate as GPS though. It will either show up as a very accurate position (no drift in position over time), or as one with a very high estimated error.

The rough location is mostly used as an input to the initial GPS equation which allows the chip to "solve" quicker. It used to be called "AGPS".

Remember back in the day a GPS would take 10 minutes to "warm up" and then it would obtain a "lock" -- phones don't need to because they start with a rough location most of the time.

The "selected availability" was turned off years ago.
What you are seeing is the actual variability due to atmospheric conditions, signal bounces, and other factors.

In fact our recent solar flares will have made all of your GPS units a little less accurate over the past few months because it will have interfered with the signal. It's caused some havok in the sat phone area recently.

If you are going to compare GPS error stats from different units you will need to do the tests away from structures, trees or rock faces. An open field would be best. If you were to collect data in the forest or next to a building the variability caused by multipath (reflected) signals for a given unit could be greater than the difference in error between two units and your findings will be inconclusive. I once got to a job site and found that they had set up their GPS base station next to a buidling with a steel roof - it explained a number of problems seen in the data. It had to be moved to eliminate the problem.

@magnetite that's a very good point.

In my experiment I used my front porch as it was raining a little, and I live in ValleyCliffe, so the chief and Slhanay might be responsible for certain behaviours I've seen in my GPS units (phone, wilderness GPS and automobile GPS).

In a community project like this there's no true way to "control" for all of these factors -- the multipath being one we can influence by changing locations. Weather is another, and space weather (solar flare activity) is another. So it's not a true scientific experiment in this respect.

However, if one user takes two brands of GPS units and tests them at almost the same time and in the same location, they can compare them to each other. This is valuable to give you insight into how the units behave in the conditions where you're using them, which is the real point of the experiment, and why I thought users should log in to contribute: so they can compare their own GPS units.

If you upload a test, just note the conditions under the "description" field, and I'll be adding a feature to list just your tests.

he just wants to know where we live []

My suggestion -- lets meet at Boundry Bay Dyke Trail at the end of 72nd Street. There is a MASCOT marker located there, and everyone can put their GPS unit around the marker, and wait
..bring your coffee and stories.

The marker is very exact lat/long already known.
49 3 34.25276 (+/-0.007m) Long 123 1 27.07193 (+/-0.006m )

We can wait an hour or two, and see what the results are

http://a100.gov.bc.ca/pub/mascotw/public/index.html

What is the practical application of all this?

When you're in the mountains, away from trees and rock faces, you will know if it's likely you are 10 or 25 meters away from where you think you are.

When you look at your position in the GPS you'll be confident you are where it says you are because the pixel accuracy will be far below the GPS accuracy.

When you're tired of climbing mountains you can play with your tool.

People think the devices are more accurate than they are. 99% of the GPS units I've used report the 50% confidence interval, which is misleading. This project lets you compare the precision across devices, and under different environmental conditions.

For instance, the people flying UAVs keep claiming that they are accurate to within a metre, which is clearly not true, and you can see it for yourself.

When you are measuring precision using the smartphone, does the orientation matter? Is there a difference between a flat orientation and a portrait orientation similar to how you might hold it in your hand?

Is there a way to correlate the smartphone location with a base station to verify the positional accuracy? Is there an android app for that?

Thanks!